1. Field of the Invention
The present invention relates generally to the field of autostereoscopic displays. More particularly, it concerns a system and method for displaying segmented, autostereoscopic images such that they appear to form a large and continuous autostereoscopic image.
2. Description of Related Art
Autostereoscopic images, such as holograms, stereograms, holographic stereograms, lenticular images, and raster-barrier or parallax-barrier images, are two-dimensional (2D) or three-dimensional (3D) visual displays that may portray more visual information than standard 2D images, such as photographs, video displays, or monitor displays.
Due to limitations in the size of available media for recording autostereoscopic images and limitations in the autostereoscopic imaging and development technologies, it has been difficult to produce large autostereoscopic images. Typically, the largest size of available recording media is a roll with a width of about 1 or 2 meters and a length of about a few hundred meters. Thus, the width of available recording media limits one dimension of a conventional autostereoscopic image to about 1 or 2 meters. Besides the size of available recording media, other factors further limit the size of autostereoscopic images. For instance, in the case of holography, while exposing recording material, the material must not be allowed to move or vibrate more than a fraction of a wavelength. It is very difficult to hold large sheets of recording material that still while properly exposing the recording material. Some have enlisted the aid of computers to form component autostereoscopic images; however, due to limitations in computer processor speeds for the computations necessary to form computer-aided component images, the computation periods for calculating the components for a large, autostereoscopic image have been prohibitively expensive and prohibitively time consuming. Moreover, due to stringent geometrical display requirements, component images with significant depth are extremely troublesome to align. Thus, due to limitations in the size of available recording material and limitations in autostereoscopic imaging and development techniques, the largest, continuous, autostereoscopic images have generally been approximately 1 meter by approximately 1 or 2 meters.
One technique that has been used to produce large, segmented, autostereoscopic displays has been to create a multitude of separate autostereoscopic images on separate recording sheets and then to display them next to each other. This technique has allowed large displays of smaller but discontinuous images or discontinuous windows of images. Such discontinuous images or windows have been easily displayed, because the segmented images or windows are separate and do not need to be closely matched with neighboring images or windows. This technique has not been used to form a continuous autostereoscopic image or a continuous window of an autostereoscopic image from segments of recording material, because it is very difficult to a match segment of recording material with its neighboring segments such that the resulting image or window appears continuous. Moreover, aligning segments of recording material for a large, 3D autostereoscopic image is slow, troublesome, and costly.
Unlike the process for aligning segments of recording material to form a large, standard 2D image, the process for aligning segments of recording material to form a 3D autostereoscopic image has even more requirements. Where the registration error is defined as the difference between where the point of an image should be and where it is, all images require registration errors between neighboring image points on consecutive segments to translate to be less than one arc minute (1/60.degree.) relative to a viewpoint in order to appear continuous to the human eye. This requirement is readily met for 2D images, especially for large, standard 2D images such as standard 2D billboards. For standard 2D images, the visible effect of registration errors, such as those due to a gap between segments, become smaller and less noticeable with increased distance. Thus, aligning standard 2D images is easier when the image is meant to be viewed from a great distance. Moreover, 2D images are mainly sensitive to rotations about an axis perpendicular to the plane in which the image lies (i.e., the z-axis, in an orthogonal system where the x and y axes lie in the plane of the image). In other words, for 2D images small rotations about the x and y axes will not have as noticeable of an effect as small rotations about the z-axis. In contrast, many autostereoscopic images are sensitive to rotations about all three orthogonal axes. For instance, if the appropriate surface planes of the recording material of consecutive segments of the recording material for a continuous, 3D, autostereoscopic image are not aligned (i.e., are rotated about the x-axis with respect to each other) one segment may appear to be displaying a set of image points or a view that is higher or lower than the set of image points or the view displayed by a segment next to it. In addition, if the appropriate surface planes of the recording material of consecutive segments of the recording material for a continuous, 3D, autostereoscopic image are not aligned (i.e. are rotated about the y-axis with respect to each other), one segment may appear to be displaying a set of image points or the view that is more to the right or to the left of the set of image points or a view displayed by a segment next to it. It is difficult to avoid registration errors due to such angular offsets of segments about the x or y axis (i.e., tilting) because it is difficult to construct a surface, especially a large surface of rigid lightweight material, that is planar enough to avoid such offsets. Moreover, unlike standard 2D images, the registration error for a 3D autostereoscopic image linearly increases with a 3D autostereoscopic image point's distance to a recording surface. This can be a dramatic problem for a large, 3D, autostereoscopic image with significant depth, because the deeper the image, the more sensitive the display is to registration errors.